[pull] main from Alishahryar1:main

.github/workflows/tests.yml

@@ -173,6 +173,7 @@ jobs:
       - name: Run Playwright
         run: >-
           uv run --locked pytest e2e -v --tb=short
+          -o faulthandler_timeout=10 --durations=20
           --tracing=retain-on-failure --screenshot=only-on-failure
           --full-page-screenshot --output=test-results
 

README.md

@@ -25,7 +25,7 @@
 
 ## What You Get
 
-- **54 ToS-friendly providers. 1.3B+ free tokens every month.** Use free, paid, subscription, and local models from one searchable UI without putting your account at risk. FCC follows provider terms and removes integrations if they stop being allowed.
+- **55 ToS-friendly providers. 1.3B+ free tokens every month.** Use free, paid, subscription, and local models from one searchable UI without putting your account at risk. FCC follows provider terms and removes integrations if they stop being allowed.
 - **10 coding agents. One model catalog.** Run [Claude Code](https://code.claude.com/docs/en/overview), [Codex](https://github.com/openai/codex), [Pi](https://github.com/earendil-works/pi), [OpenCode](https://github.com/anomalyco/opencode), [Cline](https://github.com/cline/cline), [Hermes](https://github.com/NousResearch/hermes-agent), [DeepSeek Harness](https://github.com/deepseek-ai/deepseek-harness), [Grok Build](https://github.com/xai-org/grok-build), [Muse Code](https://research.meta.ai/blog/introducing-muse-code-and-muse-spark-1-2/), or [Aider](https://aider.chat/) with your FCC models.
 - **Keep coding through provider outages.** After retries are exhausted, FCC automatically tries your next configured model without making you restart the turn. This works across every client.
 - **Up to 90% fewer terminal-output tokens.** Optional [RTK](https://github.com/rtk-ai/rtk) filters common command output, while five FCC optimizations handle quota probes, command-prefix detection, titles, suggestions, and filepaths without calling a provider.
@@ -253,6 +253,7 @@ from more than one provider before succeeding.
 | [Scaleway](https://console.scaleway.com/iam/api-keys) | `SCW_SECRET_KEY` | `scaleway/deepseek/deepseek-v4-flash` |
 | [Lightning AI](https://lightning.ai/) | `LIGHTNING_API_KEY` | `lightning/lightning-ai/Qwen3.8-27B` |
 | [Experiential Labs](https://platform.experientiallabs.ai/) | `EXPLABS_API_KEY` | `experiential/union-alpha` |
+| [Cheaper Inference](https://cheaperinference.com/signup) | `CHEAPER_INFERENCE_API_KEY` | `cheaperinference/gpt-5.4-mini` |
 | [Ollama Cloud](https://ollama.com/settings/keys) | `OLLAMA_API_KEY` | `ollama_cloud/qwen3-coder:480b` |
 | [LM Studio](https://lmstudio.ai/) | `LM_STUDIO_BASE_URL` | `lmstudio/<model-id>` |
 | [llama.cpp](https://github.com/ggml-org/llama.cpp) | `LLAMACPP_BASE_URL` | `llamacpp/<model-id>` |

e2e/conftest.py

@@ -1,6 +1,7 @@
 """Isolated browser-test composition for the local Admin UI."""
 
 import asyncio
+import faulthandler
 import socket
 import sys
 import threading
@@ -348,6 +349,7 @@ def run_server() -> None:
         listener.close()
         clear_settings_cache()
         if thread.is_alive():
+            faulthandler.dump_traceback(file=sys.stderr)
             pytest.fail("Admin browser-test server did not stop")
 
 

e2e/test_connected_accounts.py

@@ -1,5 +1,7 @@
 """Rendered connected-account flows use provider-owned capabilities and identity."""
 
+import time
+
 import pytest
 from playwright.sync_api import Dialog, Page, Route, expect
 
@@ -55,6 +57,15 @@ def __init__(self, base_url: str) -> None:
         self.hold_login = False
         self.pending_login: list[Route] = []
 
+    def take_pending_status(self, page: Page) -> Route:
+        deadline = time.monotonic() + 5
+        while not self.pending_status:
+            if time.monotonic() >= deadline:
+                pytest.fail("Account status request did not reach the route handler")
+            # Pump Playwright events so the route handler can enqueue the request.
+            page.wait_for_timeout(10)
+        return self.pending_status.pop(0)
+
     def config(self, route: Route) -> None:
         response = route.fetch()
         config = response.json()
@@ -163,19 +174,14 @@ def _open(page: Page, admin_base_url: str) -> None:
 def test_account_modes_wait_for_status_and_recover_after_load_failure(
     page: Page, admin_base_url: str, accounts: _Accounts
 ) -> None:
-    page.expose_function(
-        "pendingAccountStatusCount", lambda: len(accounts.pending_status)
-    )
     accounts.hold_status.add("github_copilot")
     page.goto(f"{admin_base_url}/admin")
     copilot = page.locator('[data-provider="github_copilot"]')
     openai = page.locator('[data-provider="openai"]')
     expect(copilot.get_by_role("button", name="Loading…", exact=True)).to_be_disabled()
     expect(copilot.get_by_role("button", name="Connect", exact=True)).to_have_count(0)
     expect(openai.get_by_role("button", name="Connect", exact=True)).to_be_enabled()
-    page.wait_for_function("async () => await window.pendingAccountStatusCount() === 1")
-
-    accounts.pending_status.pop().fulfill(
+    accounts.take_pending_status(page).fulfill(
         status=503, json={"detail": "Account status unavailable."}
     )
     expect(copilot.locator(".provider-meta")).to_have_text(
@@ -184,9 +190,9 @@ def test_account_modes_wait_for_status_and_recover_after_load_failure(
     expect(copilot.get_by_role("button", name="Connect", exact=True)).to_have_count(0)
     copilot.get_by_role("button", name="Retry", exact=True).click()
     expect(copilot.get_by_role("button", name="Loading…", exact=True)).to_be_disabled()
-    page.wait_for_function("async () => await window.pendingAccountStatusCount() === 1")
+    retry = accounts.take_pending_status(page)
     accounts.hold_status.clear()
-    accounts.pending_status.pop().fulfill(json=accounts.statuses["github_copilot"])
+    retry.fulfill(json=accounts.statuses["github_copilot"])
 
     expect(copilot.get_by_role("button", name="Connect", exact=True)).to_be_enabled()
     expect(copilot.get_by_role("button", name="Use device code")).to_have_count(0)

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "free-claude-code"
-version = "6.2.64"
+version = "6.2.66"
 description = "Local proxy connecting coding agents to OpenAI-compatible AI providers"
 readme = "README.md"
 license = "AGPL-3.0-only"

smoke/lib/config.py

@@ -95,6 +95,7 @@
     "llm7": "llm7/default",
     "lightning": "lightning/lightning-ai/Qwen3.8-27B",
     "experiential": "experiential/union-alpha",
+    "cheaperinference": "cheaperinference/gpt-5.4-mini",
     "agnes": "agnes/agnes-2.0-flash",
     "zenmux": "zenmux/deepseek/deepseek-v4-flash-free",
     "wandb": "wandb/openai/gpt-oss-20b",

src/free_claude_code/api/admin_static/providers/SOURCES.md

@@ -11,6 +11,7 @@ The LLaMA.cpp icon comes from the upstream llama.cpp repository; its MIT license
 | `azure-color.svg` | https://unpkg.com/@lobehub/icons-static-svg@1.95.0/icons/azure-color.svg |
 | `bedrock-color.svg` | https://unpkg.com/@lobehub/icons-static-svg@1.95.0/icons/bedrock-color.svg |
 | `cerebras-color.svg` | https://unpkg.com/@lobehub/icons-static-svg@1.95.0/icons/cerebras-color.svg |
+| `cheaperinference.svg` | https://www.cheaperinference.com/icon.svg |
 | `chutes.png` | https://chutes.ai/favicon.png |
 | `cline.svg` | https://unpkg.com/@lobehub/icons-static-svg@1.95.0/icons/cline.svg |
 | `cloudflare-color.svg` | https://unpkg.com/@lobehub/icons-static-svg@1.95.0/icons/cloudflare-color.svg |

src/free_claude_code/config/admin/provider_manifest.py

@@ -297,6 +297,14 @@ class ProviderFieldOverride(TypedDict, total=False):
             "hex characters; mint one at platform.experientiallabs.ai/settings/api-keys."
         ),
     },
+    "CHEAPER_INFERENCE_API_KEY": {
+        "label": "Cheaper Inference API Key",
+        "description": (
+            "Cheaper Inference OpenAI-compatible gateway API key for "
+            "api.cheaperinference.com/v1. "
+            "Keys start with ci_live_; create one at cheaperinference.com/signup."
+        ),
+    },
 }
 
 

src/free_claude_code/config/provider_catalog.py

@@ -86,6 +86,8 @@
 LIGHTNING_DEFAULT_BASE = "https://lightning.ai/api/v1"
 # Experiential Labs OpenAI-compatible Chat Completions gateway.
 EXPERIENTIAL_DEFAULT_BASE = "https://api.experientiallabs.ai/v1"
+# Cheaper Inference OpenAI-compatible Chat Completions gateway.
+CHEAPERINFERENCE_DEFAULT_BASE = "https://api.cheaperinference.com/v1"
 # Agnes AI OpenAI-compatible Chat Completions API.
 AGNES_DEFAULT_BASE = "https://apihub.agnes-ai.com/v1"
 # ZenMux OpenAI-compatible Chat Completions gateway.
@@ -687,6 +689,17 @@ def configuration_attrs(self) -> tuple[str, ...]:
         default_base_url=EXPERIENTIAL_DEFAULT_BASE,
         proxy_attr="experiential_proxy",
     ),
+    "cheaperinference": ProviderDescriptor(
+        provider_id="cheaperinference",
+        display_name="Cheaper Inference",
+        website_url="https://cheaperinference.com/",
+        logo_filename="cheaperinference.svg",
+        credential_env="CHEAPER_INFERENCE_API_KEY",
+        credential_url="https://cheaperinference.com/signup",
+        credential_attr="cheaperinference_api_key",
+        default_base_url=CHEAPERINFERENCE_DEFAULT_BASE,
+        proxy_attr="cheaperinference_proxy",
+    ),
     "ollama_cloud": ProviderDescriptor(
         provider_id="ollama_cloud",
         display_name="Ollama Cloud",

src/free_claude_code/config/settings.py

@@ -195,6 +195,11 @@ class Settings(BaseModel):
         default=None, validation_alias="EXPLABS_API_KEY"
     )
 
+    # ==================== Cheaper Inference (OpenAI-compatible) ====================
+    cheaperinference_api_key: OptionalNonEmptyString = Field(
+        default=None, validation_alias="CHEAPER_INFERENCE_API_KEY"
+    )
+
     # ==================== Fireworks AI Config ====================
     fireworks_api_key: OptionalNonEmptyString = Field(
         default=None, validation_alias="FIREWORKS_API_KEY"
@@ -498,6 +503,9 @@ class Settings(BaseModel):
     experiential_proxy: OptionalNonEmptyString = Field(
         default=None, validation_alias="EXPLABS_PROXY"
     )
+    cheaperinference_proxy: OptionalNonEmptyString = Field(
+        default=None, validation_alias="CHEAPER_INFERENCE_PROXY"
+    )
     fireworks_proxy: OptionalNonEmptyString = Field(
         default=None, validation_alias="FIREWORKS_PROXY"
     )

src/free_claude_code/core/anthropic/tokens.py

@@ -1,15 +1,30 @@
 """Token estimation for Anthropic-compatible requests."""
 
 import json
+from typing import Any
 
 from loguru import logger
 
 from free_claude_code.core.token_estimation import estimate_text_tokens
 
-from .content import get_block_attr
+from .content import get_block_attr, get_block_type
 from .models import Message, SystemContent, Tool
 
 
+def _image_tokens(block: Any) -> int:
+    """Estimate one image block from its base64 size, or a flat cost without data.
+
+    Tool-result content is not validated, so ``data`` can be any JSON value there;
+    only a string has a meaningful size, anything else gets the flat cost.
+    """
+    source = get_block_attr(block, "source")
+    if isinstance(source, dict):
+        data = source.get("data") or source.get("base64") or ""
+        if isinstance(data, str) and data:
+            return max(85, len(data) // 3000)
+    return 765
+
+
 def get_token_count(
     messages: list[Message],
     system: str | list[SystemContent] | None = None,
@@ -50,20 +65,21 @@ def get_token_count(
                     total_tokens += estimate_text_tokens(str(block_id))
                     total_tokens += 15
                 elif b_type == "image":
-                    source = get_block_attr(block, "source")
-                    if isinstance(source, dict):
-                        data = source.get("data") or source.get("base64") or ""
-                        if data:
-                            total_tokens += max(85, len(data) // 3000)
-                        else:
-                            total_tokens += 765
-                    else:
-                        total_tokens += 765
+                    total_tokens += _image_tokens(block)
                 elif b_type == "tool_result":
                     content = get_block_attr(block, "content", "")
                     tool_use_id = get_block_attr(block, "tool_use_id", "")
                     if isinstance(content, str):
                         total_tokens += estimate_text_tokens(content)
+                    elif isinstance(content, list):
+                        # Images returned by a tool (Read on an image, MCP screenshots)
+                        # are priced like top-level images; serializing them would
+                        # count their base64 payload as text.
+                        images = [p for p in content if get_block_type(p) == "image"]
+                        others = [p for p in content if get_block_type(p) != "image"]
+                        total_tokens += sum(_image_tokens(image) for image in images)
+                        if others or not images:
+                            total_tokens += estimate_text_tokens(json.dumps(others))
                     else:
                         total_tokens += estimate_text_tokens(json.dumps(content))
                     total_tokens += estimate_text_tokens(str(tool_use_id))

src/free_claude_code/providers/credential_validation.py

@@ -156,6 +156,7 @@ class _Probe:
     # https://inference-docs.cerebras.ai/api-reference/models/list-models
     # https://docs.sambanova.ai/docs/api-reference/models/get-environments-available-model-list-metadata
     # https://docs.fireworks.ai/api-reference/list-accounts
+    # https://cheaperinference.com/docs
     _Probe("deepinfra", "https://api.deepinfra.com/v1/me", _identity("uid")),
     _Probe("mistral", "/models", _MODELS),
     _Probe("wandb", "/models", _MODELS),
@@ -167,6 +168,7 @@ class _Probe:
     ),
     _Probe("cerebras", "/models", _MODELS),
     _Probe("sambanova", "/models", _MODELS),
+    _Probe("cheaperinference", "/models", _MODELS),
     _Probe(
         "fireworks",
         "https://api.fireworks.ai/v1/accounts?pageSize=1",

src/free_claude_code/providers/openai_chat/profiles.py

@@ -723,6 +723,24 @@ def _zai_profile(provider_name: str) -> OpenAIChatProfile:
         ),
         model_listing=OpenAIModelListing(path="/models"),
     ),
+    "cheaperinference": OpenAIChatProfile(
+        _policy(
+            "CHEAPERINFERENCE",
+            ReasoningReplayMode.REASONING_CONTENT,
+            default_max_tokens=ANTHROPIC_DEFAULT_MAX_OUTPUT_TOKENS,
+        ),
+        NamedEffortReasoning(
+            _LOW_MEDIUM_HIGH,
+            enabled_value="medium",
+        ),
+        model_listing=OpenAIModelListing(
+            path="/models",
+            query_params=(("type", "text"), ("streaming", "true")),
+            required_path_values=((("type",), ("text",)),),
+            context_window_tokens_path=("context_length",),
+            max_output_tokens_path=("max_output_tokens",),
+        ),
+    ),
     "ollama_cloud": OpenAIChatProfile(
         _policy(
             "OLLAMA_CLOUD",

tests/api/test_request_utils.py

@@ -1,5 +1,6 @@
 """Tests for API request detection and token counting helpers."""
 
+import json
 from unittest.mock import MagicMock, patch
 
 import pytest
@@ -549,6 +550,78 @@ def test_image_block_with_dict_source(self):
         count = get_token_count([msg])
         assert count >= 85
 
+    def test_image_in_tool_result_costs_the_same_as_a_top_level_image(self):
+        """A tool-returned image (Read on a PNG, an MCP screenshot) is priced as an image.
+
+        Serializing it with the rest of the tool result counted its base64 payload as
+        text: a 300 KB screenshot came out near 215k tokens instead of about 100.
+        """
+        image = {
+            "type": "image",
+            "source": {
+                "type": "base64",
+                "media_type": "image/png",
+                "data": "A" * 300_000,
+            },
+        }
+        text = {"type": "text", "text": "Screenshot captured"}
+
+        def user_message(*blocks):
+            msg = MagicMock()
+            msg.content = list(blocks)
+            return msg
+
+        def tool_result(*parts):
+            return user_message(
+                {"type": "tool_result", "tool_use_id": "t1", "content": list(parts)}
+            )
+
+        def image_cost(wrap):
+            return get_token_count([wrap(text, image)]) - get_token_count([wrap(text)])
+
+        assert image_cost(tool_result) == image_cost(user_message)
+        assert image_cost(tool_result) < 1_000
+
+    @pytest.mark.parametrize("data", [12345, ["a"], {"x": 1}])
+    def test_tool_result_image_with_non_string_data_gets_the_flat_cost(self, data):
+        """Tool-result content is not validated, so image data can be any JSON value.
+
+        A number used to reach ``len()`` and turn count_tokens into an HTTP 500.
+        """
+        image = {"type": "image", "source": {"type": "base64", "data": data}}
+        msg = MagicMock()
+        msg.content = [{"type": "tool_result", "tool_use_id": "t1", "content": [image]}]
+        no_data = MagicMock()
+        no_data.content = [
+            {
+                "type": "tool_result",
+                "tool_use_id": "t1",
+                "content": [{"type": "image", "source": {"type": "base64"}}],
+            }
+        ]
+
+        assert get_token_count([msg]) == get_token_count([no_data])
+
+    @pytest.mark.parametrize(
+        "parts",
+        [
+            [{"type": "text", "text": "line one"}, {"type": "text", "text": "two"}],
+            [],
+        ],
+    )
+    def test_image_free_tool_result_list_is_counted_as_before(self, parts):
+        """Only images leave the serialized payload; image-free results are unchanged."""
+        msg = MagicMock()
+        msg.content = [{"type": "tool_result", "tool_use_id": "t1", "content": parts}]
+
+        with patch(
+            "free_claude_code.core.anthropic.tokens.estimate_text_tokens",
+            side_effect=len,
+        ):
+            count = get_token_count([msg])
+
+        assert count == len(json.dumps(parts)) + len("t1") + 8 + 4
+
     def test_known_payload_structural_overhead(self):
         """Protocol overhead remains separate from plain-text estimation."""
         system_text = "You are a helpful assistant."

tests/contracts/test_provider_catalog_order.py

@@ -56,6 +56,7 @@
     "scaleway",
     "lightning",
     "experiential",
+    "cheaperinference",
     "ollama_cloud",
     "lmstudio",
     "llamacpp",